Abstract: On the basis of image fusion and feature transform for ASTER original 9 bands, in combination with the training data of 8 land cover types distributed within the study area and the measures of 48 sample plots of poplar, objective evaluation in terms of performance of classification and correlation of modeling was conducted in this paper. The results indicated that combination of vegetation indices with the highest average separability was taken as the optimal subset for subsequent classification to better characterize their spatial distribution of predominant land cover types. Meanwhile, H feature derived from HIS fusion possessed a strong correlation with ground average height and stem volume of poplar, and NDVI played a significant role in retrieving age of poplar.